det.mlir.config.yaml

---

name: PP-OCRv3_det

gops: [5.558682168]
shapes:
  - [1, 3, 960, 608]

mlir_transform:
  model_transform.py 
    --model_name $(name) 
    --model_def $(home)/ch_PP-OCRv3_det_infer.onnx 
    --input_shapes [[1,3,960,608]] 
    --mean 0.0,0.0,0.0 
    --scale 0.0039216,0.0039216,0.0039216 
    --keep_aspect_ratio 
    --pixel_format rgb 
    --output_names sigmoid_0.tmp_0 
    --test_input $(root)/dataset/samples/dog.jpg 
    --test_result $(workdir)/$(name)_top_outputs.npz 
    --mlir $(workdir)/$(name).mlir

#run_calibration.py ch_PP-OCRv3_det_infer.mlir \
#  --dataset ../ILSVRC2012 \
#  --input_num 100 \
#  -o det_cali_table

BM1684X:
  deploy:
    - model_deploy.py 
        --mlir $(workdir)/$(name).mlir 
        --quantize F32 
        --chip bm1684x 
        --test_input $(workdir)/$(name)_in_f32.npz 
        --test_reference $(workdir)/$(name)_top_outputs.npz 
        --model $(workdir)/$(name)_1684x_f32.bmodel

#  - model_deploy.py 
#      --mlir $(workdir)/$(name).mlir 
#      --quantize INT8 
#      --calibration_table $(home)/det_cali_table 
#      --chip bm1684x 
#      --test_input $(workdir)/$(name)_in_f32.npz 
#      --test_reference $(workdir)/$(name)_top_outputs.npz 
#      --tolerance 0.85,0.45 
#      --model $(workdir)/$(name)_1684x_int8_sym.bmodel